pacman::p_load(sf, tmap, tidyverse)Hands-on_Ex07
1 Choropleth Mapping with R
Choropleth mapping involves the symbolisation of enumeration units, such as countries, provinces, states, counties or census units, using area patterns or graduated colors. For example, a social scientist may need to use a choropleth map to portray the spatial distribution of aged population of Singapore by Master Plan 2014 Subzone Boundary.
In this chapter, I will plot functional and truthful choropleth maps by using an R package called tmap package.
1.1 Data Loading
The first data set to be used for below is called SGPools_svy21. The data is in csv file format.
For the second part of this exercise, two data sets will be used below to create the choropleth map:
Master Plan 2014 Subzone Boundary (Web) (i.e. MP14_SUBZONE_WEB_PL) in ESRI shapefile format. It can be downloaded at data.gov.sg. This is a geospatial data. It consists of the geographical boundary of Singapore at the planning subzone level. The data is based on URA Master Plan 2014.
Singapore Residents by Planning Area / Subzone, Age Group, Sex and Type of Dwelling, June 2011-2020 in csv format (i.e. respopagesextod2011to2020.csv). This is an aspatial data fie. It can be downloaded at Department of Statistics, Singapore Although it does not contain any coordinates values, but it’s PA and SZ fields can be used as unique identifiers to geocode to MP14_SUBZONE_WEB_PL shapefile.
mpsz <- st_read(dsn = "Data/geospatial",
layer = "MP14_SUBZONE_WEB_PL")Reading layer `MP14_SUBZONE_WEB_PL' from data source
`G:\My Drive\SMU MITB\ISSS608 Visual Analytics and Applications\lzy66666666\Visual-Analytics-Applications\Hands-on_Ex\Ex07\Data\geospatial'
using driver `ESRI Shapefile'
Simple feature collection with 323 features and 15 fields
Geometry type: MULTIPOLYGON
Dimension: XY
Bounding box: xmin: 2667.538 ymin: 15748.72 xmax: 56396.44 ymax: 50256.33
Projected CRS: SVY21popdata <- read_csv("Data/aspatial/respopagesextod2011to2020.csv")1.2 Data Wrangling
popdata2020 <- popdata %>%
filter(Time == 2020) %>%
group_by(PA, SZ, AG) %>%
summarise(`POP` = sum(`Pop`)) %>%
ungroup()%>%
pivot_wider(names_from=AG,
values_from=POP) %>%
mutate(YOUNG = rowSums(.[3:6])
+rowSums(.[12])) %>%
mutate(`ECONOMY ACTIVE` = rowSums(.[7:11])+
rowSums(.[13:15]))%>%
mutate(`AGED`=rowSums(.[16:21])) %>%
mutate(`TOTAL`=rowSums(.[3:21])) %>%
mutate(`DEPENDENCY` = (`YOUNG` + `AGED`)
/`ECONOMY ACTIVE`) %>%
select(`PA`, `SZ`, `YOUNG`,
`ECONOMY ACTIVE`, `AGED`,
`TOTAL`, `DEPENDENCY`)Before we can perform the georelational join, one extra step is required to convert the values in PA and SZ fields to uppercase. This is because the values of PA and SZ fields are made up of upper- and lowercase. On the other, hand the SUBZONE_N and PLN_AREA_N are in uppercase.
popdata2020 <- popdata2020 %>%
mutate_at(.vars = vars(PA, SZ),
.funs = funs(toupper)) %>%
filter(`ECONOMY ACTIVE` > 0)Next, left_join() of dplyr is used to join the geographical data and attribute table using planning subzone name e.g. SUBZONE_N and SZ as the common identifier.
mpsz_pop2020 <- left_join(mpsz, popdata2020,
by = c("SUBZONE_N" = "SZ"))left_join() of dplyr package is used with mpsz simple feature data frame as the left data table is to ensure that the output will be a simple features data frame.
write_rds(mpsz_pop2020, "data/rds/mpszpop2020.rds")Two approaches can be used to prepare thematic map using tmap, they are:
Plotting a thematic map quickly by using qtm().
Plotting highly customisable thematic map by using tmap elements.
1.3 Plotting a choropeth map quickly by using qtm()
The easiest and quickest to draw a choropleth map using tmap is using qtm(). It is concise and provides a good default visualisation in many cases.
tmap_mode("plot")
qtm(mpsz_pop2020,
fill = "DEPENDENCY")
Take note that:
tmap_mode() with “plot” option is used to produce a static map. For interactive mode, “view” option should be used.
fill argument is used to map the attribute (i.e. DEPENDENCY)
1.4 Creating a choropleth map by using tmap’s elements
Despite its usefulness of drawing a choropleth map quickly and easily, the disadvantge of qtm() is that it makes aesthetics of individual layers harder to control. To draw a high quality cartographic choropleth map as shown in the figure below, tmap’s drawing elements should be used.
tm_shape(mpsz_pop2020)+
tm_fill("DEPENDENCY",
style = "quantile",
palette = "Blues",
title = "Dependency ratio") +
tm_layout(main.title = "Distribution of Dependency Ratio by planning subzone",
main.title.position = "center",
main.title.size = 1.2,
legend.height = 0.45,
legend.width = 0.35,
frame = TRUE) +
tm_borders(alpha = 0.5) +
tm_compass(type="8star", size = 2) +
tm_scale_bar() +
tm_grid(alpha =0.2) +
tm_credits("Source: Planning Sub-zone boundary from Urban Redevelopment Authorithy (URA)\n and Population data from Department of Statistics DOS",
position = c("left", "bottom"))
1.4.0.1 Drawing a base map
The basic building block of tmap is tm_shape() followed by one or more layer elemments such as tm_fill() and tm_polygons().
In the code chunk below, tm_shape() is used to define the input data (i.e mpsz_pop2020) and tm_polygons() is used to draw the planning subzone polygons
tm_shape(mpsz_pop2020) +
tm_polygons()
1.4.0.2 Drawing a choropleth map using tm_polygons()
To draw a choropleth map showing the geographical distribution of a selected variable by planning subzone, we just need to assign the target variable such as Dependency to tm_polygons().
tm_shape(mpsz_pop2020)+
tm_polygons("DEPENDENCY")
Take note that:
The default interval binning used to draw the choropleth map is called “pretty”. A detailed discussion of the data classification methods supported by tmap will be provided in sub-section later.
The default colour scheme used is YlOrRd of ColorBrewer. By default, Missing value will be shaded in grey.
1.4.0.3 Drawing a choropleth map using tm_fill and tm_border()
tm_polygons() is a wraper of tm_fill() and tm_border(). tm_fill() shades the polygons by using the default colour scheme and tm_borders() adds the borders of the shapefile onto the choropleth map.
tm_shape(mpsz_pop2020)+
tm_fill("DEPENDENCY")
tm_shape(mpsz_pop2020)+
tm_fill("DEPENDENCY") +
tm_borders(lwd = 0.1, alpha = 1)
Take note that:
The alpha argument is used to define transparency number between 0 (totally transparent) and 1 (not transparent). By default, the alpha value of the col is used (normally 1).
Beside alpha argument, there are three other arguments for tm_borders(), they are:
col = border colour,
lwd = border line width. The default is 1, and
lty = border line type. The default is “solid”.
1.4.0.4 Data classification methods of tmap
Most choropleth maps employ some methods of data classification. The point of classification is to take a large number of observations and group them into data ranges or classes.
tmap provides a total ten data classification methods, namely: fixed, sd, equal, pretty (default), quantile, kmeans, hclust, bclust, fisher, and jenks.
To define a data classification method, the style argument of tm_fill() or tm_polygons() will be used.
1.4.0.4.1 Plotting choropleth maps with built-in classification methods
A quantile data classification that used 5 classes.
tm_shape(mpsz_pop2020)+
tm_fill("DEPENDENCY",
n = 5,
style = "quantile") +
tm_borders(alpha = 0.5)
An equal data classification method
tm_shape(mpsz_pop2020)+
tm_fill("DEPENDENCY",
n = 5,
style = "equal") +
tm_borders(alpha = 0.5)
Take note that from above plots, the distribution of quantile data classification method are more evenly distributed then equal data classification method.
1.4.0.5 Plotting choropleth map with customised break
For all the built-in styles, the category breaks are computed internally. In order to override these defaults, the breakpoints can be set explicitly by means of the breaks argument to the tm_fill(). It is important to note that, in tmap the breaks include a minimum and maximum. As a result, in order to end up with n categories, n+1 elements must be specified in the breaks option (the values must be in increasing order).
Before getting started, it is always a good practice to get some descriptive statistics on the variable before setting the break points. Code chunk below will be used to compute and display the descriptive statistics of DEPENDENCY field.
summary(mpsz_pop2020$DEPENDENCY) Min. 1st Qu. Median Mean 3rd Qu. Max. NA's
0.1111 0.7147 0.7866 0.8585 0.8763 19.0000 92 With reference to the results above, we set break point at 0.60, 0.70, 0.80, and 0.90. In addition, we also need to include a minimum and maximum, which we set at 0 and 100. Our breaks vector is thus c(0, 0.60, 0.70, 0.80, 0.90, 1.00)
Now, we will plot the choropleth map by using the code chunk below.
tm_shape(mpsz_pop2020)+
tm_fill("DEPENDENCY",
breaks = c(0, 0.60, 0.70, 0.80, 0.90, 1.00)) +
tm_borders(alpha = 0.5)
1.4.0.6 Color Scheme
tmap supports colour ramps either defined by the user or a set of predefined colour ramps from the RColorBrewer package.
1.4.0.6.1 Using ColourBrewer palette
To change the colour, we assign the preferred colour to palette argument of tm_fill() as shown in the code chunk below.
tm_shape(mpsz_pop2020)+
tm_fill("DEPENDENCY",
n = 6,
style = "quantile",
palette = "Blues") +
tm_borders(alpha = 0.5)
We may reverse the color shading by adding a prefix of “-” in front of the color.
tm_shape(mpsz_pop2020)+
tm_fill("DEPENDENCY",
n = 6,
style = "quantile",
palette = "-Blues") +
tm_borders(alpha = 0.5)
1.4.1 Map Layouts
Map layout refers to the combination of all map elements into a cohensive map. Map elements include among others the objects to be mapped, the title, the scale bar, the compass, margins and aspects ratios. Colour settings and data classification methods covered in the previous section relate to the palette and break-points are used to affect how the map looks.
1.4.1.1 Map legend
In tmap, several legend options are provided to change the placement, format and appearance of the legend.
tm_shape(mpsz_pop2020)+
tm_fill("DEPENDENCY",
style = "jenks",
palette = "Blues",
legend.hist = TRUE,
legend.is.portrait = TRUE,
legend.hist.z = 0.1) +
tm_layout(main.title = "Distribution of Dependency Ratio by planning subzone \n(Jenks classification)",
main.title.position = "center",
main.title.size = 1,
legend.height = 0.45,
legend.width = 0.35,
legend.outside = FALSE,
legend.position = c("right", "bottom"),
frame = FALSE) +
tm_borders(alpha = 0.5)
1.4.1.2 Map style
tmap allows a wide variety of layout settings to be changed. They can be called by using tmap_style().
E.g. classic style
tm_shape(mpsz_pop2020)+
tm_fill("DEPENDENCY",
style = "quantile",
palette = "-Greens") +
tm_borders(alpha = 0.5) +
tmap_style("classic")
1.4.1.3 Catographic furniture
Beside map style, tmap also also provides arguments to draw other map furniture such as compass, scale bar and grid lines.
In the code chunk below, tm_compass(), tm_scale_bar() and tm_grid() are used to add compass, scale bar and grid lines onto the choropleth map.
tm_shape(mpsz_pop2020)+
tm_fill("DEPENDENCY",
style = "quantile",
palette = "Blues",
title = "No. of persons") +
tm_layout(main.title = "Distribution of Dependency Ratio \nby planning subzone",
main.title.position = "center",
main.title.size = 1.2,
legend.height = 0.45,
legend.width = 0.35,
frame = TRUE) +
tm_borders(alpha = 0.5) +
tm_compass(type="8star", size = 2) +
tm_scale_bar(width = 0.15) +
tm_grid(lwd = 0.1, alpha = 0.2) +
tm_credits("Source: Planning Sub-zone boundary from Urban Redevelopment Authorithy (URA)\n and Population data from Department of Statistics DOS",
position = c("left", "bottom"))
1.4.1.4 Drawing Small Multiple CHoropleth Maps
Small multiple maps, also referred to as facet maps, are composed of many maps arrange side-by-side, and sometimes stacked vertically. Small multiple maps enable the visualisation of how spatial relationships change with respect to another variable, such as time.
In tmap, small multiple maps can be plotted in three ways:
by assigning multiple values to at least one of the asthetic arguments,
by defining a group-by variable in tm_facets(), and
by creating multiple stand-alone maps with tmap_arrange().
1.4.1.4.1 By assigning multiple values to at least one of the aesthetic arguments
tm_shape(mpsz_pop2020)+
tm_fill(c("YOUNG", "AGED"),
style = "equal",
palette = "Blues") +
tm_layout(legend.position = c("right", "bottom")) +
tm_borders(alpha = 0.5) +
tmap_style("white")
1.4.1.4.2 By defining a group-by variable in tm_facets()
tm_shape(mpsz_pop2020) +
tm_fill("DEPENDENCY",
style = "quantile",
palette = "Blues",
thres.poly = 0) +
tm_facets(by="REGION_N",
free.coords=TRUE,
drop.shapes=TRUE) +
tm_layout(legend.show = FALSE,
title.position = c("center", "center"),
title.size = 20) +
tm_borders(alpha = 0.5)
1.4.1.4.3 By creating multiple stand-alone maps with tmap_arrange()
youngmap <- tm_shape(mpsz_pop2020)+
tm_polygons("YOUNG",
style = "quantile",
palette = "Blues")
agedmap <- tm_shape(mpsz_pop2020)+
tm_polygons("AGED",
style = "quantile",
palette = "Blues")
tmap_arrange(youngmap, agedmap, asp=1, ncol=2)
1.4.1.5 Mappping Spatial Object Meeting a Selection Criterion
Instead of creating small multiple choropleth map, you can also use selection funtion to map spatial objects meeting the selection criterion.
tm_shape(mpsz_pop2020[mpsz_pop2020$REGION_N=="CENTRAL REGION", ])+
tm_fill("DEPENDENCY",
style = "quantile",
palette = "Blues",
legend.hist = TRUE,
legend.is.portrait = TRUE,
legend.hist.z = 0.1) +
tm_layout(legend.outside = TRUE,
legend.height = 0.45,
legend.width = 5.0,
legend.position = c("right", "bottom"),
frame = FALSE) +
tm_borders(alpha = 0.5)
2 Visualising Geospatial Point Data
Proportional symbol maps (also known as graduate symbol maps) are a class of maps that use the visual variable of size to represent differences in the magnitude of a discrete, abruptly changing phenomenon, e.g. counts of people. Like choropleth maps, you can create classed or unclassed versions of these maps. The classed ones are known as range-graded or graduated symbols, and the unclassed are called proportional symbols, where the area of the symbols are proportional to the values of the attribute being mapped. In this hands-on exercise, you will learn how to create a proportional symbol map showing the number of wins by Singapore Pools’ outlets using an R package called tmap.
pacman::p_load(sf, tmap, tidyverse)2.1 Data Wangling
The data set used for this hands-on exercise is called SGPools_svy21. The data is in csv file format.
Figure below shows the first 15 records of SGPools_svy21.csv. It consists of seven columns. The XCOORD and YCOORD columns are the x-coordinates and y-coordinates of SingPools outlets and branches.
sgpools <- read_csv("data/aspatial/SGPools_svy21.csv")list(sgpools) [[1]]
# A tibble: 306 × 7
NAME ADDRESS POSTC…¹ XCOORD YCOORD OUTLE…² Gp1Gp…³
<chr> <chr> <dbl> <dbl> <dbl> <chr> <dbl>
1 Livewire (Marina Bay Sands) 2 Bayf… 18972 30842. 29599. Branch 5
2 Livewire (Resorts World Sentos… 26 Sen… 98138 26704. 26526. Branch 11
3 SportsBuzz (Kranji) Lotus … 738078 20118. 44888. Branch 0
4 SportsBuzz (PoMo) 1 Sele… 188306 29777. 31382. Branch 44
5 Prime Serangoon North Blk 54… 552542 32239. 39519. Branch 0
6 Singapore Pools Woodlands Cent… 1A Woo… 731001 21012. 46987. Branch 3
7 Singapore Pools 64 Circuit Rd … Blk 64… 370064 33990. 34356. Branch 17
8 Singapore Pools 88 Circuit Rd … Blk 88… 370088 33847. 33976. Branch 16
9 Singapore Pools Anchorvale Rd … Blk 30… 540308 33910. 41275. Branch 21
10 Singapore Pools Ang Mo Kio N2 … Blk 20… 560202 29246. 38943. Branch 25
# … with 296 more rows, and abbreviated variable names ¹POSTCODE,
# ²`OUTLET TYPE`, ³`Gp1Gp2 Winnings`We convert sgpools data frame into a simple feature data frame by using st_as_sf() of sf packages
sgpools_sf <- st_as_sf(sgpools,
coords = c("XCOORD", "YCOORD"),
crs= 3414)The coords argument requires you to provide the column name of the x-coordinates first then followed by the column name of the y-coordinates.
The crs argument required you to provide the coordinates system in epsg format. EPSG: 3414 is Singapore SVY21 Projected Coordinate System. You can search for other country’s epsg code by refering to epsg.io.
2.2 Drawing Proportional Symbol Map
To create an interactive proportional symbol map in R, the view mode of tmap will be used.
tmap_mode("view")tm_shape(sgpools_sf) +
tm_bubbles(col = "red",
size = 1,
border.col = "black",
border.lwd = 1)2.2.1 Attributes with different colors
tm_shape(sgpools_sf)+
tm_bubbles(col = "OUTLET TYPE",
size = "Gp1Gp2 Winnings",
border.col = "black",
border.lwd = 1)2.2.2 tm_facets() to produce multiple maps with synchronised zoom and pan settings
tm_shape(sgpools_sf) +
tm_bubbles(col = "OUTLET TYPE",
size = "Gp1Gp2 Winnings",
border.col = "black",
border.lwd = 1) +
tm_facets(by= "OUTLET TYPE",
nrow = 1,
sync = TRUE)3 Analytical Mapping
3.1 Packages and data loading
pacman::p_load(tmap, tidyverse, sf)NGA_wp <- read_rds("data/rds/NGA_wp.rds")3.2 Visualising distribution of non-functional water point
p1 <- tm_shape(NGA_wp) +
tm_fill("wp_functional",
n = 10,
style = "equal",
palette = "Blues") +
tm_borders(lwd = 0.1,
alpha = 1) +
tm_layout(main.title = "Distribution of functional water point by LGAs",
legend.outside = FALSE)p2 <- tm_shape(NGA_wp) +
tm_fill("total_wp",
n = 10,
style = "equal",
palette = "Blues") +
tm_borders(lwd = 0.1,
alpha = 1) +
tm_layout(main.title = "Distribution of total water point by LGAs",
legend.outside = FALSE)tmap_arrange(p2, p1, nrow = 1)3.3 Choropleth Map for Rates
In much of our readings we have now seen the importance to map rates rather than counts of things, and that is for the simple reason that water points are not equally distributed in space. That means that if we do not account for how many water points are somewhere, we end up mapping total water point size rather than our topic of interest.
3.3.1 Deriving Proportion of Functional Water Points and Non-Functional Water Points
We will tabulate the proportion of functional water points and the proportion of non-functional water points in each LGA. In the following code chunk, mutate() from dplyr package is used to derive two fields, namely pct_functional and pct_nonfunctional.
NGA_wp <- NGA_wp %>%
mutate(pct_functional = wp_functional/total_wp) %>%
mutate(pct_nonfunctional = wp_nonfunctional/total_wp)3.3.2 Plotting map of rate
tm_shape(NGA_wp) +
tm_fill("pct_functional",
n = 10,
style = "equal",
palette = "Blues",
legend.hist = TRUE) +
tm_borders(lwd = 0.1,
alpha = 1) +
tm_layout(main.title = "Rate map of functional water point by LGAs",
legend.outside = TRUE)3.4 Extreme value maps
Extreme value maps are variations of common choropleth maps where the classification is designed to highlight extreme values at the lower and upper end of the scale, with the goal of identifying outliers. These maps were developed in the spirit of spatializing EDA, i.e., adding spatial features to commonly used approaches in non-spatial EDA (Anselin 1994).
3.4.1 Percentile Map
The percentile map is a special type of quantile map with six specific categories: 0-1%,1-10%, 10-50%,50-90%,90-99%, and 99-100%. The corresponding breakpoints can be derived by means of the base R quantile command, passing an explicit vector of cumulative probabilities as c(0,.01,.1,.5,.9,.99,1). Note that the begin and endpoint need to be included.
Step 1: Exclude records with NA by using the code chunk below.
NGA_wp <- NGA_wp %>%
drop_na()Step 2: Creating customised classification and extracting values
percent <- c(0,.01,.1,.5,.9,.99,1)
var <- NGA_wp["pct_functional"] %>%
st_set_geometry(NULL)
quantile(var[,1], percent) 0% 1% 10% 50% 90% 99% 100%
0.0000000 0.0000000 0.2169811 0.4791667 0.8611111 1.0000000 1.0000000 Creating the get.var function
Firstly, we will write an R function as shown below to extract a variable (i.e. wp_nonfunctional) as a vector out of an sf data.frame.
arguments:
vname: variable name (as character, in quotes)
df: name of sf data frame
returns: v: vector with values (without a column name)
get.var <- function(vname,df) {
v <- df[vname] %>%
st_set_geometry(NULL)
v <- unname(v[,1])
return(v)
}3.4.1.1 A percentile mapping function
percentmap <- function(vnam, df, legtitle=NA, mtitle="Percentile Map"){
percent <- c(0,.01,.1,.5,.9,.99,1)
var <- get.var(vnam, df)
bperc <- quantile(var, percent)
tm_shape(df) +
tm_polygons() +
tm_shape(df) +
tm_fill(vnam,
title=legtitle,
breaks=bperc,
palette="Blues",
labels=c("< 1%", "1% - 10%", "10% - 50%", "50% - 90%", "90% - 99%", "> 99%")) +
tm_borders() +
tm_layout(main.title = mtitle,
title.position = c("right","bottom"))
}percentmap("total_wp",
NGA_wp)3.4.2 Box map
In essence, a box map is an augmented quartile map, with an additional lower and upper category. When there are lower outliers, then the starting point for the breaks is the minimum value, and the second break is the lower fence. In contrast, when there are no lower outliers, then the starting point for the breaks will be the lower fence, and the second break is the minimum value (there will be no observations that fall in the interval between the lower fence and the minimum value).
boxbreaks <- function(v,mult=1.5) {
qv <- unname(quantile(v))
iqr <- qv[4] - qv[2]
upfence <- qv[4] + mult * iqr
lofence <- qv[2] - mult * iqr
# initialize break points vector
bb <- vector(mode="numeric",length=7)
# logic for lower and upper fences
if (lofence < qv[1]) { # no lower outliers
bb[1] <- lofence
bb[2] <- floor(qv[1])
} else {
bb[2] <- lofence
bb[1] <- qv[1]
}
if (upfence > qv[5]) { # no upper outliers
bb[7] <- upfence
bb[6] <- ceiling(qv[5])
} else {
bb[6] <- upfence
bb[7] <- qv[5]
}
bb[3:5] <- qv[2:4]
return(bb)
}get.var <- function(vname,df) {
v <- df[vname] %>% st_set_geometry(NULL)
v <- unname(v[,1])
return(v)
}var <- get.var("wp_nonfunctional", NGA_wp)
boxbreaks(var)[1] -56.5 0.0 14.0 34.0 61.0 131.5 278.0boxmap <- function(vnam, df,
legtitle=NA,
mtitle="Box Map",
mult=1.5){
var <- get.var(vnam,df)
bb <- boxbreaks(var)
tm_shape(df) +
tm_polygons() +
tm_shape(df) +
tm_fill(vnam,title=legtitle,
breaks=bb,
palette="Blues",
labels = c("lower outlier",
"< 25%",
"25% - 50%",
"50% - 75%",
"> 75%",
"upper outlier")) +
tm_borders() +
tm_layout(main.title = mtitle,
title.position = c("left",
"top"))
}tmap_mode("plot")
boxmap("wp_nonfunctional", NGA_wp)